Hi @jonradoff
First — incredible work. The scope of what you shipped here (multi-tenancy, white-label branding, MCP server, health monitoring, outgoing webhooks, dual credit buckets...) is genuinely impressive, and the fact that it was built entirely through Claude Code makes it even more interesting to me.
I'm curious about the how behind the conversation-driven development:
- How did you structure your prompts? Did you give Claude Code a full spec upfront, or did you iterate feature by feature?
- Did you use a system prompt or CLAUDE.md file to give the agent context about the codebase as it grew?
- How did you handle consistency across such a large codebase — did the agent ever drift from established patterns, and how did you course-correct?
- Any example prompts you'd be willing to share? Even a rough sense of the style (e.g. "Add a webhook system that supports X, Y, Z with HMAC signing") would be really valuable.
Hi @jonradoff
First — incredible work. The scope of what you shipped here (multi-tenancy, white-label branding, MCP server, health monitoring, outgoing webhooks, dual credit buckets...) is genuinely impressive, and the fact that it was built entirely through Claude Code makes it even more interesting to me.
I'm curious about the how behind the conversation-driven development: